Title: Identifying Drug Mechanism of Action using Network Dysregulation Abstract: Identifying the Mechanism of Action (MOA) of drugs is critical for the development of a new drug, understanding the drug side effects, and repositioning the existing drugs. However, interrogating drug MoA has been challenging since it requires both understanding activities of all the molecules participating interactions in the biological network, and a well-designed reverse engineering methodology to understand the information flow among the network. In this study, we developed an algorithm based upon the two key advances to identify molecular targets of drugs and MOAs, which are high-throughput gene expression technology and systems biology approaches. The overall procedure comprises by three steps. First, we a construct a network including protein-protein interactions and DNA-protein interactions using ARACNe, an information theory based reverse engineering algorithm. We then calculate the significance of dysregulation edges by measuring the difference between distributions of a certain phenotype and controls in two-dimensional space using Gaussian kernel method and Kullback–Leibler (KL) divergence. Once measuring the significance of edges in the entire network is done, we score the nodes by integration of p-values from edges surrounding the node, the p-value that represents the significance of dysregulation of each edge. The nodes enriched by dysregulated edges will be ranked high in the final results. To test the ability to capture a drug target in the perturbed network, we generated a huge compendium of dataset composed of 288 microarray gene expression profiles from samples treated by 14 drugs. In an analysis of the dataset, we could compare the performance on predicting known targets of 10 drugs by comparing their ranks among 2,0071 genes. We observed that our approach outperforms the conventional differential gene analysis. The performance also held on many of drug treated datasets downloaded from the public gene expression datasets. Based on our method, we could have better picture for a network perturbed by a certain compound. The next step would be applying this methodology for better understanding on drug off-target events, response rate of drugs, or the possibility of repositioning of an existing drug.